Fix NVMe hot-swap LED stuck in FAILURE under VMD controllers

src/ledmon/ledmon.c

@@ -670,6 +670,14 @@ static void _add_block(struct block_device *block)
 			} else {
 				temp->ibpi = block->ibpi;
 			}
+		} else if (temp->ibpi == LED_IBPI_PATTERN_FAILED_DRIVE &&
+			   !temp->raid_dev) {
+			/*
+			 * Non-RAID device reappeared in sysfs scan.
+			 * Allow recovery from FAILED_DRIVE so the LED
+			 * is cleared after a hot-swap cycle.
+			 */
+			temp->ibpi = LED_IBPI_PATTERN_ADDED;
 		} else if (!(temp->ibpi == LED_IBPI_PATTERN_FAILED_DRIVE &&
 			block->ibpi == LED_IBPI_PATTERN_HOTSPARE) ||
 			(temp->ibpi == LED_IBPI_PATTERN_FAILED_DRIVE &&

src/ledmon/udev.c

@@ -27,24 +27,40 @@ static struct udev_monitor *udev_monitor;
 
 static int _compare(const struct block_device *bd, const char *syspath, struct led_ctx *ctx)
 {
+	struct block_device *bd_new;
+	const char *udev_name;
+	const char *dev_name;
+	int ret;
+
 	if (!bd || !syspath)
 		return 0;
 
-	if (strcmp(bd->sysfs_path, syspath) == 0) {
+	if (strcmp(bd->sysfs_path, syspath) == 0)
 		return 1;
-	} else {
-		struct block_device *bd_new;
-		int ret;
-
-		bd_new = block_device_init(sysfs_get_cntrl_devices(ctx), syspath);
-		if (!bd_new)
-			return 0;
 
+	bd_new = block_device_init(sysfs_get_cntrl_devices(ctx), syspath);
+	if (bd_new) {
 		ret = block_compare(bd, bd_new);
 		block_device_fini(bd_new);
-
-		return ret;
+		if (ret)
+			return 1;
 	}
+
+	/*
+	 * NVMe udev events may arrive with a virtual nvme-subsystem path
+	 * that cannot be resolved to a PCI controller. Fall back to matching
+	 * by device name so that add/remove events are not silently dropped.
+	 */
+	if (bd->devnode[0] == '\0')
+		return 0;
+
+	dev_name = strrchr(bd->devnode, '/');
+	dev_name = dev_name ? dev_name + 1 : bd->devnode;
+
+	udev_name = strrchr(syspath, '/');
+	udev_name = udev_name ? udev_name + 1 : syspath;
+
+	return strcmp(dev_name, udev_name) == 0;
 }
 
 static int create_udev_monitor(void)

src/lib/pci_slot.c

@@ -5,6 +5,7 @@
 #include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
+#include <sys/stat.h>
 #include <unistd.h>
 #include "config.h"
 #include "led/libled.h"
@@ -71,6 +72,12 @@ struct slot_property *pci_slot_property_init(struct pci_slot *pci_slot)
 
 	result->bl_device = get_block_device_from_sysfs_path(pci_slot->ctx,
 							     pci_slot->address, true);
+	if (result->bl_device) {
+		struct stat st;
+
+		if (stat(result->bl_device->devnode, &st) != 0)
+			result->bl_device = NULL;
+	}
 	result->slot_spec.pci = pci_slot;
 	snprintf(result->slot_id, PATH_MAX, "%s", pci_slot->sysfs_path);
 	result->c = &pci_slot_common;